Physician, Practice, and Patient Characteristics Associated With Biosimilar Use in Medicare Recipients

Key Points Question What patient, physician, and practice characteristics are associated with biosimilar usage for the biologics filgrastim and infliximab? Findings In this cross-sectional study of 40 656 Medicare fee-for-service beneficiaries, few patient and physician characteristics were associated with biosimilar usage. Practice setting characteristics had the largest associations; however, the types of practices with high biosimilar use differed by drug class. Meaning In this study, practice setting and hospital ownership status were associated with use of biosimilars, but further research is needed to understand the reasons for differences across drug classes.


Introduction
Biologic medicines-complex medicines derived from biologic sources-account for a growing share of drug spending. 1 Biologic medicine spending in the United States increased by 50.1% 2 between 2014 and 2018, putting additional financial pressure on public and private payers. In 2010, the Biologics Price Competition and Innovation Act created a pathway for the development, approval, and launch of highly similar biosimilar medications to compete with originator biologic medications. 3 Similar to the role of generic medicines for small-molecule drugs, biosimilars are intended as a substitute for biologic drugs, thus creating competition in the biologics market to drive spending down.
Biosimilars have the potential to decrease drug spending growth; however, products have been slow to launch in the United States. As of June 30, 2020, 27 biosimilars have been approved and 14 have launched, compared with 58 approvals in Europe. 4,5 Even when launched, biosimilar uptake in the United States has been slow. [6][7][8][9] Identifying characteristics associated with biosimilar use can clarify which barriers are likely driving the slow adoption [10][11][12][13][14] and may help to develop policies to encourage the uptake of biosimilars.
This study asks what patient, physician, and practice characteristics are associated with biosimilar usage for the biologics filgrastim and infliximab in the Medicare fee-for-service population through 2018. Two previous studies examined characteristics associated with filgrastim biosimilar uptake in the Medicare fee-for-service population through 2017 15 and in the commercial and Medicare Advantage settings through 2018. 9 Two studies 6,16 have examined uptake of infliximab through 2018 in the Medicare fee-for-service population. This study extends previous analysis for the filgrastim drug class in the Medicare fee-for-service population and is the first to examine associations with practice characteristics and a wider range of physician characteristics in the infliximab drug class.

Data and Sample
The primary data source was claims and enrollment data from 2014 to 2018 for a random 20% sample of beneficiaries enrolled in Parts A and B of fee-for-service Medicare. All Part B biosimilar administrations were identified for filgrastim products (filgrastim, filgrastim-sndz, and close competitor tbo-filgrastim) and for infliximab products (infliximab, infliximab-dyyb, and infliximab- This study was deemed exempt from review by the institutional review board at Weill Cornell Medical College, including a waiver of informed consent due to the use of secondary data. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies. 17

JAMA Network Open | Health Policy
Physician, Practice, and Patient Characteristics Associated With Biosimilar Use

Outcomes and Covariates
The primary outcome was a binary variable indicating whether a Part B biologic administration was a biosimilar. Patient characteristics included indicators for age group (ie, 65-74 or Ն75 years), sex, race, dual Medicare-Medicaid eligibility, relevant medical conditions, and risk scores. Risk scores were based on the Health and Human Services Hierarchical Conditions Categories risk adjustment model using claims from the preceding year. 14 For patients receiving biosimilars, we identified whether a patient had previously used the originator biologic. Characteristics were present at the time of a patient's first administration (biosimilar or originator biologic) in the sample.
The administering physician was identified on a beneficiary's biologic claim (eAppendix in the Supplement). Physician characteristics included age, sex, and indicators for physician specialty from the 2018 Physician Compare database. Additionally, indicators for whether a physician practiced in a hospital-owned practice and low, medium, or high biologic administration volume were constructed from Medicare claims ( Table 1). If a physician only appeared in years prior to 2018, physician characteristics were collected from the relevant year of Physician Compare data. Finally, a physician's percentage of biosimilar prescriptions during the full sample period was calculated for physicians who administered any biosimilars.
Biosimilars may be administered in the hospital outpatient department setting or a physician's office. Practice characteristics were measured separately for each setting. In the hospital outpatient setting, characteristics included indicators for hospital size (<50, 50-100, and Ն100 beds), ownership (not-for-profit, for-profit, and government), academic medical center status and system affiliation from the 2018 American Hospital Association Survey, and 340B status in 2018 from the Health Resources and Services Administration Office of Pharmacy Affairs. In the office setting, characteristics included indicators for practice size (1-5, 6-19, 20-99, and Ն100 physicians) and multispecialty status ( Table 2). 18 The percentage of biosimilar prescriptions during the full sample was calculated for hospital outpatient departments or offices that administered at least 1 biosimilar medication.

Statistical Analysis
We conducted a number of analyses to examine the extent to which patient, physician, and practice characteristics differed across administrations of biosimilars and originator biologics. Analyses were conducted separately for each drug class.
Mean-comparison t tests or tests of proportion were used to compare patient, physician, and practice characteristics across groups. Patient characteristics were compared across patients who ever received a biosimilar vs those who never received a biosimilar. Physician and practice characteristics were compared between physicians and practices that ever administered a biosimilar vs those that did not.
Ordinary least squares multivariable regressions were used to examine the patient, physician, and practice characteristics associated with biosimilar administration. The 3 main regressions included (1) patient and physician characteristics, with an indicator for administration in a hospital outpatient setting; (2) patient, physician, and office characteristics only for administrations within the office setting; and (3) patient, physician, and hospital outpatient characteristics only for administrations within the hospital outpatient setting. All models also controlled for quarter-year and state fixed effects. Patient risk score was excluded from the main regressions, but it was included in sensitivity analyses. In all regressions, robust standard errors were clustered at the patient level.
Sensitivity analyses included comparing patient characteristics in a sample limited to facilities that prescribed at least 1 biosimilar. To examine the sensitivity of results to model specifications, additional analyses included (1) addition of patient risk score covariate; (2) clustering at physician and separately practice level; (3) exclusion of first quarter after initial biosimilar launch; (4) restriction of filgrastim sample to first 9 quarters, the same number as the infliximab sample; (5) restriction to final year or 2 quarters of sample; (6) exclusion of beneficiaries or physicians with low volume; and (7) inclusion only of high volume physicians. The study was conducted using SAS software version 9.4 (SAS Institute Inc) and Stata statistical software version 14.1 (StataCorp). Statistical significance was set at P < .05.

Results
The final filgrastim sample included 25 870 patients ( h For filgrastim sample, low, medium, and high volume were defined as at least 1, more than 1 to 5, and more than 5 average monthly filgrastim administrations, respectively. For infliximab, low, medium, and high volume were defined as at least 1, more than 1 to 2, and more than 2 average monthly infliximab administrations, respectively.

JAMA Network Open | Health Policy
Physician, Practice, and Patient Characteristics Associated With Biosimilar Use between the quarter 4 2016 launch of infliximab-abda and quarter 4 2018 (eFigure 1 and eFigure 2 in the Supplement). Because of missing covariate data, 6725 administrations (2.5%) and 7798 (4.3%) for the filgrastim and infliximab drug classes, respectively, were excluded. Product market share and volume by month appear in eFigure 3 to eFigure 6 in the Supplement. While the market share of filgrastim-sndz has risen to 52% by December 2018, infliximab biosimilars reached 10% of the market by December 2018.
Patients who received at least 1 biosimilar had small differences in demographic and risk characteristics compared with those who had never received a biosimilar, with the exception of age for the filgrastim patients (  In adjusted filgrastim analyses (Figure 1) administered in a for-profit setting than a not-for-profit setting (P < .001).
In adjusted infliximab analyses (Figure 2), the only patient characteristic associated with a biosimilar administration was presence of Crohn disease; these patients were 1.8 (95% CI, 0.8 to 2.9)

JAMA Network Open | Health Policy
Physician, Practice, and Patient Characteristics Associated With Biosimilar Use    percentage points more likely to be a biosimilar (P < .001). Infliximab biosimilar administrations were more likely to take place in larger offices or hospital outpatient departments. Numerous other hospital outpatient setting characteristics were associated with biosimilar administrations.
Sensitivity analyses, presented in eTable 2 to eTable 9 in the Supplement, yielded similar results.

Discussion
The decision to administer a biosimilar rather than an originator biologic may be influenced by characteristics of patients, physicians, or the setting of administration. In this study of factors associated with biosimilar uptake in the Medicare fee-for-service population, patient characteristics were weakly associated with biosimilar uptake, while few physician characteristics, including higher product volume and physician specialty, were associated with biosimilar use. The practice setting of an administration had the strongest association with whether a patient received a biosimilar; however, the direction of the association differed by drug class. In adjusted analyses, a patient in the hospital outpatient setting was 16 percentage points or 42% less likely to receive a filgrastim biosimilar than a patient in an office setting but 3 percentage points or 73% more likely to receive an infliximab biosimilar. For administrations in a hospital outpatient setting, ownership status was also associated with biosimilar utilization but again in opposite directions by drug class. Patients were 17 percentage points less likely and 11 percentage points more likely to receive filgrastim and infliximab biosimilars, respectively, in for-profit vs not-for-profit hospital outpatient departments.
This work supports previous evidence from studies of filgrastim, which suggest that patient characteristics are not strongly associated with biosimilar use. 15 Adjusted analyses controlling for patient and practice characteristics found that both physician specialty 15,16 and physician prescribing volume 15 of a given drug class remained strongly associated with biosimilar use. Although results were not consistent in the filgrastim drug class, physicians with a higher volume of prescriptions were generally more likely to prescribe biosimilar products. This could be because of limited awareness or comfort with biosimilars among physicians with lower rates of prescribing. 19 How to encourage use of biosimilars among this latter group of physicians is an open question.
To our knowledge, this is the first study to document that practice setting is associated with infliximab biosimilar administrations, with uptake significantly higher in the hospital outpatient setting than the physician office setting. The association in practice setting uptake was inverted for filgrastim; there was a higher rate of biosimilar uptake in physician practices than in hospital outpatient departments, a result that was consistent with previous work. Importantly, this study demonstrates that factors associated with biosimilar use may differ across drug classes, and therefore, each drug class may require different interventions to promote use.
The fact that practice setting was consistently associated with prescribing when controlling for patient and physician characteristics suggests that physician practices or hospital outpatient departments may play an important role in steering prescribers toward certain medications. This could occur either explicitly or implicitly, for instance by limiting products on a formulary or purchasing schedule. Practices and hospital outpatient departments may want to steer physicians toward certain products when they can earn higher profits on those products. For the average practice and hospital outpatient department, the profit from administering a biologic and biosimilar to a Medicare beneficiary is the same, given that Medicare reimburses the medication at their respective average sales price plus 6% of the reference biologic's average sales price. As originator

JAMA Network Open | Health Policy
Physician, Practice, and Patient Characteristics Associated With Biosimilar Use biologics are generally more expensive than biosimilars, it is possible that practices that receive the largest discounts may find originator biologics more profitable if they are able to get a lower net price for these products. Traditionally, hospital outpatient departments are able to receive larger discounts relative to smaller office practices due to their high-volume purchasing, which could be an explanation for slower filgrastim-sndz uptake in the hospital outpatient setting. However, this fact may not hold for infliximab: the manufacturer of the originator (Johnson and Johnson) has pursued unique contracting mechanisms to discourage biosimilar usage, which could have a large financial impact on both small and large organizations.
Other proposed explanations for differential uptake between practice settings have included patients with more severe illness selecting 1 setting over another and more complicated purchasing processes in the hospital outpatient setting, leading to slower adoption. For the former, we found that practice setting remained associated with biosimilar use even when controlling for patient and physician characteristics, suggesting this is not a complete explanation. For the latter, the fact that infliximab uptake was actually higher in the hospital outpatient setting suggests that while complex purchasing agreements may play a role, they are also not the complete story.
As this work demonstrates, adoption of biosimilars in Medicare has been uneven across different products. The uptake of filgrastim-sndz has risen consistently over time, reaching 52% of the market by December 2018. However, the uptake of infliximab biosimilars has been significantly slower, reaching 10% of the market by December 2018. This pattern is consistent with the European experience, where infliximab uptake was significantly slower than filgrastim, 6 and represents a lost opportunity for cost savings. There are a number of potential explanations for the slower uptake of infliximab biosimilars. Given that infliximab is used for chronic conditions, physicians may be less willing to switch patients who are doing well on the originator brand to a biosimilar. 10 Furthermore, financial incentives for infliximab and its biosimilars are likely to differ from filgrastim. The unique contracting mechanisms used by the manufacturer of originator infliximab (ie, Johnson and Johnson) have included exclusionary contracts and purchasing bundles with health care professionals, health care systems, and insurers to allegedly block biosimilar competition. These contracting mechanisms include so-called rebate traps, which withdraw the rebates payers and clinicians get for prescribing biologics if patients switch to a biosimilar. 10 Future work should investigate how contracting mechanisms might encourage, rather than discourage, the use of more efficient alternatives, such as biosimilars.

Limitations
This study has several limitations. First, it focuses on the Medicare population and thus misses individuals who are privately insured, for whom uptake patterns could be different. Second, this study cannot determine whether there is a causal link between patient, physician, and practice characteristics and biosimilar administration, although the analysis controls for many observable confounders. Third, this study focuses on the short-term uptake of 3 biosimilars in 2 drug classes. It is possible that long-term trends as well as trends for biosimilars in other drug classes will differ.

Conclusions
In this cross-sectional study of factors associated with the uptake of the first 3 biosimilars launched in Medicare, the practice setting (ie, office vs hospital outpatient department) and hospital outpatient ownership status had the strongest association with biosimilar use. Surprisingly, the direction of the associations between practice setting and uptake was opposite for the 2 drug classes.
Further research is needed to understand the reasons underlying differences in biosimilar uptake across practice settings for the filgrastim and infliximab drug classes.

JAMA Network Open | Health Policy
Physician, Practice, and Patient Characteristics Associated With Biosimilar Use